Method and apparatus for a parallel data storage and processing server

The present invention concerns a parallel multiprocessor-multidisk storage server which offers low delays and high throughputs when accessing and processing one-dimensional and multi-dimensional file data such as pixmap images, text, sound or graphics. The invented parallel multiprocessor-multidisk storage server may be used as a server offering its services to a computer, to client stations residing on a network or to a parallel host system to which it is connected. The parallel storage server comprises (a) a server interface processor interfacing the storage system with a host computer, with a network or with a parallel computing system; (b) an array of disk nodes, each disk node being composed by one processor electrically connected to at least one disk and (c) an interconnection network for connecting the server interface processor with the array of disk nodes. Multi-dimensional data files such as 3-d images (for example tomographic images), respectively 2-d images (for example scanned aerial photographs) are segmented into 3-d, respectively 2-d file extents, extents being striped onto different disks. One-dimensional files are segmented into 1-d file extents. File extents of a given file may have a fixed or a variable size. The storage server is based on a parallel image and multiple media file storage system. This file storage system includes a file server process which receives from the high level storage server process file creation, file opening, file closing and file deleting commands. It further includes extent serving processes running on disk node processors, which receive from the file server process commands to update directory entries and to open existing files and from the storage interface server process commands to read data from a file or to write data into a file. It also includes operation processes responsible for applying in parallel geometric transformations and image processing operations to data read from the disks and a redundancy file creation process responsible for creating redundant parity extent files for selected data files.

Skip to:  ·  Claims  ·  References Cited  · Patent History  ·  Patent History

Claims

1. A storage and processing server computing system apparatus for offering processing services including: parallel file storage. access, and processing functions, comprising:

(a) a server interface connected by a communication means to at least one computer;
(b) disk nodes, said disk nodes each comprising a disk node processor and at least one disk:
(c) an interconnection network between said server interface and disk nodes whereby: said server interface is operable for receiving service requests from computers, further operable for analyzing, processing and decomposing service requests into storage, access and processing commands and further operable for transmitting them to said disk node processors which are operable for receiving commands, further operable for executing commands and further operable for returning results, whereby an operation is selected from the group of geometric transformations and image processing operations.

2. The storage and processing server apparatus defined in claim 1, wherein service requests from remotely located client computers comprise

(a) 1-dimensional, 2-dimensional and 3-dimensional file storage and access requests;
(b) file transformation and processing requests;
(c) redundancy file creation requests;
(d) file recovery requests; wherein
(i) file storage and access requests comprise file creating, opening, and deleting commands and file extent storage and access commands executed in parallel by disk node processors;
(ii) file transformation and processing requests are decomposed into file part transformation and processing operations executed in parallel by disk node processors;
(iii) redundancy file creation requests are executed by generating from data file extents parity file extents which are stored on different disks from the disks storing the corresponding data file extents;
(iv) file recovery requests in the case of single disk crashes are executed by reading valid data extents from the file to be recovered and redundant parity extents from its corresponding redundancy file, by computing the lacking data extents and by writing the recovered file on valid disks.

3. The storage and processing server apparatus defined in claim 2, wherein file storage and access requests comprise

(a) file window storage and access commands, decomposed into extent window storage and access commands executed in parallel by disk node processors;
(b) scaled-down file window storage and access commands, decomposed into scaled-down extent window storage and access commands executed in parallel by disk node processors;
(c) file aligned window access commands, decomposed into complete extent storage and access commands executed in parallel by disk node processors.

4. The storage and processing server apparatus defined in claim 3, where aligned window access commands are decomposed into variable size extent storage and access commands executed in parallel by disk node processors.

5. The storage and processing server apparatus defined in claim 2, wherein file transformation and processing operations comprise rotation operations executed in parallel by disk node processors.

6. A parallel file storage, access and processing method comprising the steps of

(a) receiving storage, access and processing requests from a client process;
(b) decomposing file storage, access and processing requests into a plurality of access, storage and processing requests executable by disk node processors:
(c) sending said plurality of access, storage and processing requests to disk node processors:
(d) receiving in parallel said plurality of access, storage and processing requests by disk node processors;
(e) executing in parallel said plurality of requests including an operation selected from the group of geometric transformations and image processing operations by accessing disks block on a plurality of disks and by applying processing operations executed by a plurality of processors;
(f) assembling the results obtained by the parallel execution of said requests into one result:
(g) returning said result to the requesting client process.

7. The parallel file storage, access and processing method defined in claim 6, wherein files are multidimensional files made of elementary elements, each element being composed of at least one byte, wherein multidimensional files are segmented into multidimensional extents, wherein extents are incrementally numbered as is shown in the case of the following 3-dimensional extent positioned as the ith extent in the x direction, the jth extent in y direction, and the kth extent in z direction according to the formula

DiskNumber(ExtentNumber)=ExtentNumber MODULO d,

8. A multiprocessor-multidisk storage server offering parallel file storage, access, and processing functions comprising:

(a) a server interface comprising at least one local processor connected by a communication means to at least one computer;
(b) disk nodes, said disk nodes each comprising a disk node processor and at least one disk, said disk being operable for storing data;
(c) an interconnection network between said server interface and disk nodes, whereby: said server interface is operable for receiving service requests from computers, further operable for analyzing, processing and decomposing service requests into storage, access and processing commands and further operable for transmitting them to said disk node processors which are operable for receiving commands, further operable for executing commands including an operation selected from the group of geometric transformations and image processing operations and further operable for returning results.

9. A multiprocessor multidisk storage server of claim 8, where the interconnection network between the server interface and the disk nodes comprises a crossbar switch.

10. A multiprocessor multidisk storage server of claim 8, where the number of disk nodes and the number of disks per disk node can be independently chosen according to the requirements of the client application.

11. A multiprocessor multidisk storage server of claim 8, where aligned window access commands are decomposed into variable size extent storage and access commands executed in parallel by disk node processors.

12. A multiprocessor multidisk storage server of claim 8, where file transformation and processing operations comprise rotation operations executed in parallel by disk node processors.

13. A multiprocessor multidisk storage server of claim 8 where file storage and access requests comprise:

(a) file window storage and access commands, decomposed into extent window storage and access commands executed in parallel by disk node processors;
(b) scaled-down file window storage and access commands, decomposed into scaled-down extent window storage and access commands executed in parallel by disk node processors;
(c) file aligned window access commands, decomposed into complete extent storage and access commands executed in parallel by disk node processors.

14. A parallel file storage, access and processing method using processes running on a server interface processor and processes running on disk node processing units, the method comprising the steps:

(a) receiving by the processes running on the server interface processor storage, access and processing requests from a client process;
(b) decomposing file storage, access and processing requests into a plurality of access, storage and processing requests executable by processes running on disk node processors:
(c) sending said plurality of access, storage and processing requests to processes running on disk node processors:
(d) receiving in parallel said plurality of access, storage and processing requests by processes running on disk node processors:
(e)executing in parallel said plurality of requests including an operation selected from the group of geometric transformations and image processing operations by accessing disk data on a plurality of disks and by applying processing operations executed by processes running on a plurality of processors:
(f) assembling the results obtained by the parallel execution of said requests into one result; and
(g) returning said result to the requesting client process.

15. A parallel file storage, access and processing method of claim 14, where processes running in a server interface processor comprise a server interface process and at least one file server process and where processes running on disk node processing units comprise extent server processes, said server interface process being responsible for receiving file storage, access and processing requests and for transmitting file creation, opening and closing requests to the file server process and for transmitting file access requests to the extent server processes, said file server process being responsible for interacting with the extent server processes for creating new file entries in a file directory and in a distribution index block table, said distribution index table comprising file dimensionality information, file size information and file extent size information and comprising, for each disk on which a file is stored, a pointer to its file local extent table, said file local extent table comprising for each locally stored extent of a file the address of its starting block and its effective extent size.

16. A parallel file storage, access and processing method of claim 14, where files are multidimensional files made of elementary elements, each element being composed of at least one byte, where multidimensional files are segmented into multidimensional extents, where extents are incrementally numbered as is shown in the case of the following 3-dimensional extent positioned as the i.sup.th extent in the x direction, thej.sup.th extent in y direction, and the k.sup.th extent in z direction according to the formula:

17. A multiprocessor parallel file system operable for storing and accessing in parallel data files striped onto multiple disks, said data files being striped onto the disks according to their dimensionality which ranges from 1 to 3, the multiprocessor parallel file system comprising:

(a) a file server process operable for receiving file creation, opening and closing requests and operable for interaction with extent server processes regarding the creation and deletion of directory entries and distribution information block table entries;
(b) extent server processes running on disk node processors operable for receiving file extent storage and access requests and for receiving from the file server process commands for the creation and deletion of directory entries and distribution information block table entries; and
c) parallel file system interface library procedures for file creation, file opening, file access and file closing.

18. A parallel file system of claim 17, where as a means for efficient parallel file access, the size of the extents striped onto the disk at storage time is computed by varying the extent width so as to make the offset between the respective disk numbers of an extent lying one extent row below another extent prime in respect to the number of disks.

19. A multiprocessor-multidisk storage server offering parallel file storage, access, and processing functions comprising:

(a) a server interface processor connected by a communication means to at least one computer, said server interface processor being operable for receiving service requests from computers, further operable for analyzing, processing and decomposing service requests into storage, access and processing commands and further operable for transmitting them to disk node processors;
(b) disk nodes, said disk nodes each comprising a disk node processor and at least one disk, said disk node processor being operable for receiving commands, further operable for executing processing commands comprising operations selected from the group of geometric transformations and image processing operations;
(c) an interconnection network:between said server interface and disk nodes.

20. A multiprocessor parallel file system operable for storing, accessing and operating in parallel on data files striped onto multiple disks, the multiprocessor parallel file system comprising:

(a) a file server process operable for receiving file creation, opening and closing requests and operable for communicating with extent server processes regarding the creation and deletion of directory entries and distribution information block table entries;
(b) extent server processes running on disk node processors operable for receiving file extent storage and access requests and for receiving from the file server process commands for the creation and deletion of directory entries and of distribution information block table entries;
(c) operation slave processes running on disk node processors operable for receiving file reading commands and file operation commands, for executing them and for returning results;
(d) an operation master process operable for receiving file access and file operating requests, operable for decomposing said file access and file operating requests into file reading commands and into file operating commands, for transmitting said commands to operation slave processes and for receiving the results from said operation slave processes;
(e) a server interface process running on a server interface processor, the server interface process being operable for receiving service requests from client computers, further operable for analyzing, processing and decomposing service requests into file storage, access and file operating commands and for transmitting them to the file server, extent server and operation master processes.

21. A multiprocessor-multidisk storage sever offering parallel file storage, access, and processing functions comprising:

(a) a server interface comprising at least one local processor connected by a communication means to at least one computer;
(b) disk nodes, said disk nodes each comprising a disk node processor and at least one disk, said disk being operable for storing data;
(c) an interconnection network between said server interface and disk nodes whereby said server interface is operable for receiving service requests from computers, further operable for analyzing, processing and decomposing service requests into storage, access and processing commands and further operable for transmitting them to said disk node processors which are operable for receiving commands, further operable for executing commands and further operable for returning results; and where service requests from remotely located client computers comprise:
(A) file storage and access requests, where the dimesionality of the file is selected from The group of 1-dimension, 2-dimensions and 3-dimensions and where extents are striped onto the disks at file storage time according to The dimensionality of the file;
(B) file transformation and processing requests; wherein:
(i) file storage and access requests comprise file creating, opening, anti,deleting commands and file extent storage and access commands executed in parallel by disk node processors;
(ii) file transformation and processing requests are decomposed into file part transformation anti processing operations executed in parallel by disk node processors.

22. A multiprocessor multidisk storage server of claim 21 where service requests further comprise:

(a) redundancy file creation requests;
(b) file recovery requests; wherein
(i) redundancy file creation requests are executed by generating from data file extents parity file extents which are stored on different disks from the disks storing the corresponding data file extents;
(ii) file recovery requests in the case of single disk crashes are executed by reading valid data extents from the file robe recovered and redundant parity extents from its corresponding redundancy file, by computing the lacking data extents and by writing the recovered file on valid disks.

23. A multiprocessor multidisk storage server of claim 21 where, as a means for efficient parallel file access, the size of the extents striped onto the disk at storage time is computed by varying the extent width so as to make the offset between the respective disk numbers of an extent lying one extent row below another extent prime in respect to the number of disks.

24. A parallel fie storage, access and processing method using processes running on a server interface processor and processes running on disk node processing units, the method comprising the steps of:

(a) receiving by the processes running on the server interface processor storage, access and processing requests from a client process;
(b) decomposing file storage, access and processing requests into a plurality of access, storage and processing requests executable by processes running on disk node processors;
(c) sending said plurality of access storage and processing requests to processes running on disk node:processors;
(d) receiving in parallel: said plurality of access, storage and processing requests by processes running on disk node processors;
(e) executing in parallel said plurality of requests by accessing disk data on a plurality of disks and by applying processing operations executed by process as running on a plurality of processors;
(f) assembling the results obtained by the parallel execution of said requests into one result;
(g) returning said result to the requesting client process, wherein
(i) files are composed of extents, each extent comprising at least one elementary element, each elementary element being composed of at least one byte;,
(ii) decomposing file storage, access and processing requests comprises the operations of segmenting files into extents;
(iii) the processing operations executed by processes running on a plurality of processors comprise an operation selected from the group of geometric transformations and image processing operations.

25. A parallel file storage, access and processing method of claim 24, where multidimensional files are segmented into multidimensional extents, where extents are incrementally numbered as is shown in the ease of the following 3-dimensional extent positioned as the t.sup.th extent in the x direction, the J.sup.th extent in y direction, and the k.sup.th extent in z direction according to the formula:

DiskNumber(ExtentNumber)=ExtentNumber MODULO d,

26. A multiprocessor parallel file system operable for storing, accessing and processing in parallel data files striped onto multiple disks, said data files being striped onto the disks according to their dimensionality which ranges from 1 to 3, the multiprocessor parallel file system comprising:

(a) a file server process operable for receiving file creation, opening, closing and processing requests and operable for interaction with extent serve processes regarding the creation and deletion of directory entries and distribution information block table entries;
(b) extent server processes running on disk node processors operable for receiving file extent storage, access and processing operations requests and for receiving from the file server process commands for the creation and deletion of directory entries and distribution information block table entries;
(c) parallel file system interface library procedures for file creation, file opening, file access and file closing; wherein
(i) data files are composed of extents, each extent comprising at least one elementary element, each elementary element being composed of at least one byte:
(ii) the processing operations executed by extent server processes running on a plurality of processors comprise an operation selected from the group of geometric transformations and image processing operations.

27. A parallel file system of claim 24, where data files are multidimensional files segmented into multidimensional extents, where extents are incrementally numbered as is shown in the case of the following 3-dimensional extent positioned as the i.sup.th extent in the x direction, the j.sup.th extent in y direction, and the k.sup.th extent in z direction according to the formula

28. A multiprocessor parallel file system operable for storing, accessing and operating in parallel on data files striped onto multiple disks, the multiprocessor parallel file system comprising:

(a) a file server process operable for receiving file creation, opening and closing requests and operable for communicating with extent serve processes regarding the creation and deletion of directory entries and distribution information block table entries;
(b) extent server processes running on disk node processors operable for receiving file extent storage and access requests and for receiving from the file server process commands for the creation and deletion of directory entries and of distribution information block table entries;
(c) operation slave processes running on disk node processors operable for receiving file reading commands and file operation commands, for executing them and for returning results;
(d) an operation master process operable for receiving file access and file operating requests, operable for decomposing said file access and file operating requests into reading commands and into operating commands, for transmitting said commands to operation slave processes and for receiving the results from said operation slave processes; and
(e) a server interface process running on a server interface processor, the server interface process being operable for receiving service requests from client computers, further operable for analyzing, processing and decomposing service requests into file storage, access and file operating commands and for transmitting them to the file server, extent server and operation master processes; wherein:
(i) data files are composed of extents, each extent comprising at least one elementary element, each elementary element being composed of at least one byte;
(ii) decomposing file access requests comprises the operation of segmenting files into extents; and
(iii) file operation commands comprise an operation selected from the group of geometric transformations and image processing operations.

29. A parallel file storage and processing method offering high-bandwidth storage, access and processing of data files striped on a plurality of disks, comprising file creation, file reading, file writing, and file processing functions, whereby storage, access and processing requests are sent by client processes located outside the server, received by a server interface process located on the server interface processor, analyzed and decomposed into requests directed to a file server process, to extent server processes and to operation processes and where files are multidimensional files made of elementary elements, each element being composed by one or more bytes, wherein multidimensional files are segmented into multidimensional extents, wherein extents are incrementally numbered as is shown in the case of the following 3-dimensional extent positioned as the h.sup.th extent in the x direction, the J.sup.th extent in y direction, and the.sup.th extent in z direction according to the formula

DiskNumber(ExtentNumber)=ExtentNumber MODULO d,

30. A parallel file storage and processing method in claim 29, where storage, access and processing of data files striped on a plurality of disks.

Referenced Cited

U.S. Patent Documents

4985856 January 15, 1991 Kaufman et al.
5163131 November 10, 1992 Row et al.
5301310 April 5, 1994 Isman et al.
5361385 November 1, 1994 Bakalash
5377333 December 27, 1994 Nakagoshi et al.
5422987 June 6, 1995 Yamada

Other references

  • "Maximizing Performance in a Stripped Disk Array", P. A. Chen, D. A. Patterson; Proceedings IEEE International Symposium on Computer Architecture, Seattle, pp. 322-331, Jun. 1990. "Raster Rotation of Bilevel Bitmap Images", R. D. Hersch, Eurographics 85 Proceedings, (Ed. C. Vandoni), North-Holland, pp. 295-308, Jun. 1985. "Thinking Machines nears database market (Thinking Machines Corp. releases two low-end parallel processors)", Electronic News (1991), v38, n1935, p. 18(1), Oct. 1992.

Patent History

Patent number: 5737549
Type: Grant
Filed: Jan 31, 1994
Date of Patent: Apr 7, 1998
Assignee: Ecole Polytechnique Federale de Lausanne (Lausanne)
Inventors: Roger D. Hersch (Lausanne), Bernard Krummenacher (Chavannes)
Primary Examiner: Thomas G. Black
Assistant Examiner: Hosain T. Alam
Attorney: Joseph Page
Application Number: 8/188,589

Classifications

Current U.S. Class: 395/309; 395/806; 395/119; 395/505; 395/508; 395/520
International Classification: G06F 1300;